Forecasting the Production of Thai Pepper and Chili Pepper in Garut Indonesia Based on Fuzzy Time Series and X-13 ARIMA-SEATS Methods

Abstract

Time series data is a series of data  measured over a certain period of time based on fixed time intervals divided into seasonal and non-seasonal time series data. Anomaly observation that can affect the consistency of time series data are known as outliers. Outliers are often found in data that is influenced by weather, one of which is the production data of thai pepper and chili pepper that is one of the important types of vegetables that are cultivated commercially in tropical countries, Indonesia. A robust forecasting method against outliers is required in this case. The computational method developed at this time is the Fuzzy Time Series (FTS) method with the Fuzzy K-Medoids (FKM) clustering algorithm that can handle data that contains outliers. In addition, the X-13 ARIMA-SEATS method is a seasonal adjustment method that automatically detects and solves the problem of outliers and overcomes the effects of moving holiday. This study aims to forecast the monthly thai pepper and chili pepper production using both non-linear and linear methods. The result showed that FTS – FKM method give the best accuracy with smallest value of RMSE and MAPE in short forecasting period, otherwise X-13 ARIMA-SEATS give its best performance in long forecasting period.